While inference-time scaling through search has revolutionized Large Language Models, translating these gains to image generation has proven difficult. Recent attempts to apply search strategies to continuous diffusion models show limited benefits, with simple random sampling often performing best. We demonstrate that the discrete, sequential nature of visual autoregressive models enables effective search for image generation. We show that beam search substantially improves text-to-image generation, enabling a 2B parameter autoregressive model to outperform a 12B parameter diffusion model across benchmarks. Systematic ablations show that this advantage comes from the discrete token space, which allows early pruning and computational reuse, and our verifier analysis highlights trade-offs between speed and reasoning capability. These findings suggest that model architecture, not just scale, is critical for inference-time optimization in visual generation.
CAR: Controllable Autoregressive Modeling for Visual Generation
School of Electronic and Computer Engineering, Peking University(北京大学电子与计算机工程学院);Tencent Youtu Lab(腾讯优图实验室);Southern University of Science and Technology(南方科技大学);University of Washington(华盛顿大学)
Learning to Expand Images for Efficient Visual Autoregressive Modeling
University of Electronic Science and Technology of China(电子科技大学);School of Computer Science and Engineering, Central South University(中南大学计算机科学与工程学院);Xidian University(西安电子科技大学);SenseTime Research(商汤科技研究院);Shanghai Jiao Tong University(上海交通大学)
FlexVAR: Flexible Visual Autoregressive Modeling without Residual Prediction
Institute of Information Science, Beijing Jiaotong University(北京交通大学信息科学学院);University of Technology Sydney(悉尼科技大学);Meituan(美团);Georgia Institute of Technology(佐治亚理工学院)
Closing the Safety Gap: Surgical Concept Erasure in Visual Autoregressive Models
Harbin Institute of Technology, Shenzhen(哈尔滨工业大学(深圳));Tsinghua Shenzhen International Graduate School, Tsinghua University(清华大学深圳国际研究生院);Jilin University(吉林大学);Peng Cheng Laboratory(鹏城实验室);Department of Computer Science and Technology, Tsinghua University(清华大学计算机科学与技术系)
LSRS: Latent Scale Rejection Sampling for Visual Autoregressive Modeling
College of Artificial Intelligence, Nanjing University of Aeronautics and Astronautics, China(南京航空航天大学人工智能学院);MIIT Key Laboratory of Pattern Analysis and Machine Intelligence, Nanjing, China(信息产业部模式分析与机器智能重点实验室);The Key Laboratory of Brain-Machine Intelligence Technology, Ministry of Education, Nanjing, China(教育部脑机智能技术重点实验室)
VCE: Safe Autoregressive Image Generation via Visual Contrast Exploitation
Fudan University(复旦大学);Shanghai Key Lab of Intell. Info. Processing, School of CS, Fudan University(复旦大学计算机科学技术学院上海市智能信息处理重点实验室);Shanghai Collaborative Innovation Center on Intelligent Visual Computing(上海智能视觉计算协同创新中心);International Computer Science Institute(国际计算机科学研究所);UC Berkeley(加州大学伯克利分校)
FasterVAR: Plug-and-Play Acceleration for Visual Autoregressive Models
PCA Lab, VCIP, College of Computer Science, Nankai University(南开大学计算机学院、VCIP、PCA实验室);Program of Computer Science, City University of Hong Kong (Dongguan), China(香港城市大学(东莞)计算机系,中国);City University of Hong Kong, HK SAR, China(香港城市大学,香港特别行政区,中国);Mohamed bin Zayed University of Artificial Intelligence, UAE(阿联酋Mohamed bin Zayed人工智能大学);Linkoping University, Sweden(林地平大学,瑞典);PCA Lab, School of Intelligence Science and Technology, Nanjing University(南京大学智能科学与技术学院、PCA实验室);College of Artificial Intelligence, Jilin University(吉林大学人工智能学院)
VARPose: Flexible 2D Pose Densification via Visual Autoregressive Modeling for Enhanced 3D Lifting
School of Artificial Intelligence, Sun Yat-sen University(中山大学人工智能学院);The Technology Innovation Center for Collaborative Applications of Natural Resources Data in GBA, MNR(自然资源部粤港澳大湾区自然资源数据协同应用技术创新中心)
Planning and Rendering in Concert: DeepFusion of Autoregressive Layouts and Diffusion for Visual Text Generation
University of Science and Technology of China(中国科学技术大学);Central South University(中南大学);Louisiana State University(路易斯安那州立大学);Arizona State University(亚利桑那州立大学);Sun Yat-sen University(中山大学)
C2P-VAR: Continual and Compositional Personalization in Visual Autoregressive Models
Harbin Institute of Technology, Shenzhen(哈尔滨工业大学(深圳));Tsinghua Shenzhen International Graduate School, Tsinghua University(清华大学深圳国际研究生院);Peng Cheng Laboratory(鹏城实验室);South China University of Technology(华南理工大学)